advertisement

5.2 Matrix Equations 289 5.2 Matrix Equations Linear Equations. An m × n system of linear equations a11 x1 + a12 x2 + · · · + a1n xn a21 x1 + a22 x2 + · · · + a2n xn = b1 , = b2 , .. . am1 x1 + am2 x2 + · · · + amn xn = bm , ~ = ~b. Let A be the matrix of can be written as a matrix equation AX ~ be the column vector of variable names x1 , . . . , xn coefficients aij , let X ~ and let b be the column vector with components b1 , . . . , bn . Then a11 a21 a12 a22 ~ = AX am1 am2 · · · amn = x1 x2 .. . xn a11 x1 + a12 x2 + · · · + a1n xn a21 x1 + a22 x2 + · · · + a2n xn .. . = · · · a1n · · · a2n .. . am1 x1 + am2 x2 + · · · + amn xn b1 b2 .. . bn . ~ = ~b. Conversely, a matrix equation AX ~ = ~b corresponds Therefore, AX to a set of linear algebraic equations, because of reversible steps above. A system of linear equations can be represented by its variable list x1 , x2 , . . . , xn and its augmented matrix (1) a11 a21 a12 a22 · · · a1n · · · a2n .. . b1 b2 am1 am2 · · · amn bn . The augmented matrix is denoted aug(A, ~b) when the system is given as ~ = ~b. Given an augmented matrix and a variable list, the conversion AX ~ = back to a linear system of equations is made by expanding aug(A, ~b)Y ~ 0, where Y has components x1 , . . . , xn , −1. Because this expansion involves just dot products, it is possible to rapidly write out the linear system for a given augmented matrix. Hand work often contains the 290 following kind of exposition: x1 x2 ··· xn a11 a21 a12 a22 ··· ··· .. . a1n a2n b1 b2 am1 am2 ··· amn bn (2) In (2), a dot product is applied to the first n elements of each row, using the variable list written above the columns. The symbolic answer is set equal to the rightmost column’s entry, in order to recover the equations. It is usual in homogeneous systems to not write the column of zeros, but to deal directly with A instead of aug(A, 0). This convention is justified by arguing that the rightmost column of zeros is unchanged by swap, multiply and combination rules, which are defined the next paragraph. Elementary Row Operations. The three operations on equations which produce equivalent systems can be translated directly to row operations on the augmented matrix for the system. The rules produce equivalent systems, that is, the three rules neither create nor destroy solutions. Swap Two rows can be interchanged. Multiply A row can be multiplied by multiplier m 6= 0. Combination A multiple of one row can be added to a different row. Documentation of Row Operations. Throughout the display below, symbol s stands for source, symbol t for target, symbol m for multiplier and symbol c for constant. Swap swap(s,t) ≡ swap rows s and t. Multiply mult(t,m) ≡ multiply row t by m6= 0. Combination combo(s,t,c) ≡ add c times row s to row t 6= s. The standard for documentation is to write the notation next to the target row, which is the row to be changed. For swap operations, the notation is written next to the first row that was swapped, and optionally next to both rows. The notation was developed from early maple notation for the corresponding operations swaprow, mulrow and addrow, appearing in the maple package linalg. In early versions of maple, an additional argument A is used to reference the matrix for which the elementary row operation is to be applied. For instance, addrow(A,1,3,-5) 5.2 Matrix Equations 291 assumes matrix A is the object of the combination rule, which is documented in written work as combo(1,3,-5). In written work, symbol A is omitted, because A is the matrix appearing on the previous line of the sequence of steps. Later versions of maple use the LinearAlgebra package, and a single function RowOperation to accomplish the same thing. A short conversion table appears below. Hand-written swap(s,t) mult(t,c) combo(s,t,c) maple linalg swaprow(A,s,t) mulrow(A,t,c) addrow(A,s,t,c) maple LinearAlgebra RowOperation(A,[t,s]) RowOperation(A,t,c) RowOperation(A,[t,s],c) Conversion between packages can be controlled by the following function definitions, which causes the maple code to be the same regardless of which linear algebra package is used.6 Maple linalg combo:=(a,s,t,c)->addrow(a,s,t,c); swap:=(a,s,t)->swaprow(a,s,t); mult:=(a,t,c)->mulrow(a,t,c); Maple LinearAlgebra combo:=(a,s,t,c)->RowOperation(a,[t,s],c); swap:=(a,s,t)->RowOperation(a,[t,s]); mult:=(a,t,c)->RowOperation(a,t,c); macro(matrix=Matrix); RREF. The reduced row-echelon form of a matrix, or rrefEF., is specified by the following requirements. 1. Zero rows appear last. Each nonzero row has first element 1, called a leading one. The column in which the leading one appears, called a pivot column, has all other entries zero. 2. The pivot columns appear as consecutive initial columns of the identity matrix I. Trailing columns of I might be absent. The matrix (3) below is a typical rref which satisfies the preceding properties. Displayed secondly is the reduced echelon system (4) in the variables x1 , . . . , x8 represented by the augmented matrix (3). 6 The acronym ASTC is used for the signs of the trigonometric functions in quadrants I through IV. The argument lists for combo, swap, mult use the same order, ASTC, memorized in trigonometry as All Students Take Calculus. 292 (3) 1 0 0 0 0 0 0 x1 + 2x2 (4) x3 2 0 0 0 0 0 0 0 1 0 0 0 0 0 3 7 0 0 0 0 0 4 8 0 0 0 0 0 0 5 0 6 0 9 0 10 1 11 0 12 0 0 1 13 0 0 0 0 0 0 0 0 0 0 0 0 + 3x4 + 4x5 + 7x4 + 8x5 x6 + 5x7 + 9x7 + 11x7 x8 =6 = 10 = 12 = 13 Observe that (3) is an rref and (4) is a reduced echelon system. The initial 4 columns of the 7 × 7 identity matrix I appear in natural order in matrix (3); the trailing 3 columns of I are absent. If the rref of the augmented matrix has a leading one in the last column, then the corresponding system of equations then has an equation “0 = 1” displayed, which signals an inconsistent system. It must be emphasized that an rref always exists, even if the corresponding equations are inconsistent. Elimination Method. The elimination algorithm for equations (see page 185) has an implementation for matrices. A row is marked processed if either (1) the row is all zeros, or else (2) the row contains a leading one and all other entries in that column are zero. Otherwise, the row is called unprocessed. 1. Move each unprocessed row of zeros to the last row using swap and mark it processed. 2. Identify an unprocessed nonzero row having the least leading zeros. Apply the multiply rule to insure a leading one. Apply the combination rule to make all other entries zero in that column. The number of leading ones (lead variables) has been increased by one and the current column is a column of the identity matrix. 3. All other unprocessed rows have leading zeros up to and including the leading one column. Apply the swap rule to move the row into the lowest position, so that all unprocessed rows and zero rows follow, and mark the row as processed, e.g., replace the leading one by 1 . 4. Repeat steps 1–3, until all rows have been processed. Then all leading ones have been defined and the resulting matrix is in reduced row-echelon form. 5.2 Matrix Equations 293 Computer algebra systems and computer numerical laboratories use similar algorithms which produce exactly the same result, called the reduced row-echelon form of a matrix A, denoted by rref (A). Literature calls the algorithm Gauss-Jordan elimination. Two examples: rref (0) = 0 In step 2, all rows of the zero matrix 0 are zero. No changes are made to the zero matrix. In step 2, each row has a leading one. No changes are made to the identity matrix I. rref (I) = I Frame Sequence. A sequence of swap, multiply and combination steps applied to a system of equations is called a frame sequence. The viewpoint is that a camera is pointed over the shoulder of an assistant who writes the mathematics, and after the completion of each step, a photo is taken. The sequence of photo frames is the frame sequence. The First Frame is the original system and the Last Frame is the reduced row echelon system. The same terminology applies for systems A~x = ~b represented by an augmented matrix C = aug(A, ~b). The First Frame is C and the Last Frame is rref (C). Steps in a frame sequence can be documented by the notation swap(s,t), mult(t,m), combo(s,t,c), each written next to the target row t. During the sequence, consecutive initial columns of the identity, called pivot columns, are created as steps toward the rref . Trailing columns of the identity might not appear. An illustration: Frame 1: Frame 2: Frame 3: Frame 4: 1 1 0 0 2 −1 0 1 4 −1 0 2 1 1 0 1 0 0 0 0 1 0 0 0 2 −1 0 1 2 0 0 1 1 1 0 1 0 0 0 0 1 0 0 0 2 −1 0 1 1 1 0 1 2 0 0 1 0 0 0 0 1 0 0 0 1 2 −1 0 1 1 0 1 0 −2 0 −1 0 0 0 0 Original augmented matrix. combo(1,2,-1) Pivot column 1 completed. swap(2,3) combo(2,3,-2) 294 Frame 5: Frame 6: Frame 7: 1 0 0 0 0 −3 0 −1 1 1 0 1 0 −2 0 −1 0 0 0 0 1 0 0 0 0 −3 0 −1 1 1 0 1 0 1 0 1/2 0 0 0 0 1 0 0 0 0 −3 0 −1 1 0 0 1/2 0 1 0 1/2 0 0 0 0 1 0 0 0 Last Frame: 0 1 0 0 0 0 1 0 0 1/2 0 1/2 0 1/2 0 0 combo(2,1,-2) Pivot column 2 completed. mult(3,-1/2) combo(3,2,-1) combo(3,1,3) Pivot column 3 completed. rref found. Back-substitution and efficiency. The algorithm implemented in the preceding frame sequence is easy to learn, because the actual work is organized by creating pivot columns, via swap, combination and multiply. The pivot columns are initial columns of the identity. You are advised to learn the algorithm in this form, but please change the algorithm as you become more efficient at doing the steps. See the examples for illustrations. Back substitution. Computer implementations and also hand computation can be made more efficient by changing step 2 into step 2a and adding step 5. Step 2a introduces zeros below the leading one. Step 5 introduces zeros above each leading one, working from the lowest nonzero row back to the first row. Literature refers to step 5 as backsubstitution. The process of back-substitution is exactly the original elimination algorithm applied to the variable list in reverse order. Avoiding fractions. A matrix A containing only integer entries can often be put into reduced row-echelon form without introducing fractions. The multiply rule introduces fractions, so its use should be limited. It is advised that leading ones be introduced only when convenient, otherwise make the leading coefficient nonzero and positive. Divisions at the end of the computation will produce the rref . Clever use of the combination rule can sometimes create a leading one without introducing fractions. Consider the two rows 25 0 1 0 5 7 0 2 0 2 The second row multiplied by −4 and added to the first row effectively replaces the 25 by −3, whereupon adding the first row twice to the 5.2 Matrix Equations 295 second gives a leading one in the second row. The resulting rows are fraction-free. −3 0 −7 0 −3 1 0 −12 0 −4 Rank and Nullity. What does it mean, if the first column of a rref is the zero vector? It means that the corresponding variable x1 is a free variable. In fact, every column that does not contain a leading one corresponds to a free variable in the standard general solution of the system of equations. Symmetrically, each leading one identifies a pivot column and corresponds to a leading variable. The number of leading ones is the rank of the matrix, denoted rank(A). The rank cannot exceed the row dimension nor the column dimension. The column count less the number of leading ones is the nullity of the matrix, denoted nullity(A). It equals the number of free variables. Regardless of how matrix B arises, augmented or not, we have the relation variable count = rank(B) + nullity(B). ~ = ~b, then the variable count n comes from X ~ If B = aug(A, ~b) for AX and the column count of B is one more, or n + 1. Replacing the variable count by the column count can therefore lead to fundamental errors. Inverses. An efficient method to find the inverse B of a square matrix A, should it happen to exist, is to form the augmented matrix C = aug(A, I) and then read off B as the package of the last n columns of rref (C). This method is based upon the equivalence rref (aug(A, I)) = aug(I, B) if and only if AB = I. The next theorem aids not only in establishing this equivalence but also in the practical matter of testing a candidate solution for the inverse matrix. The proof is delayed to page 300. Theorem 6 (Inverse Test) If A and B are square matrices such that AB = I, then also BA = I. Therefore, only one of the equalities AB = I or BA = I is required to check an inverse. Theorem 7 (The rref Inversion Method) Let A and B denote square matrices. Then (a) If rref (aug(A, I)) = aug(I, B), then AB = BA = I and B is the inverse of A. (b) If AB = BA = I, then rref (aug(A, I)) = aug(I, B). (c) If rref (aug(A, I)) = aug(C, B) and C 6= I, then A is not invertible. The proof is delayed to page 300. 296 Finding Inverses. The method will be illustrated for the matrix 1 0 1 A = 0 1 −1 . 0 1 1 Define the first frame of the sequence to be C1 = aug(C, I), then compute the frame sequence to rref (C) as follows. 1 0 1 1 0 0 C1 = 0 1 −1 0 1 0 0 1 1 0 0 1 First Frame 1 0 1 1 0 0 1 0 C2 = 0 1 −1 0 0 0 2 0 −1 1 combo(3,2,-1) 1 0 1 1 0 0 1 0 C3 = 0 1 −1 0 0 0 1 0 −1/2 1/2 mult(3,1/2) 0 0 1 0 1 1 0 0 1/2 1/2 0 0 1 0 −1/2 1/2 C4 = 0 1 1 0 0 1 1/2 −1/2 1/2 1/2 C5 = 0 1 0 0 0 0 1 0 −1/2 1/2 combo(3,2,1) combo(3,1,-1) Last Frame The theory implies that the inverse of A is the matrix in the right half of the last frame: A−1 1 1/2 −1/2 1/2 1/2 = 0 0 −1/2 1/2 Answer Check. Let B equal the matrix of the last display, claimed to be A−1 . The Inverse Test, Theorem (6), says that we do not need to check both AB = I and BA = I. It is enough to check one of them. Details: AB = = = 1 0 0 1 0 0 1 0 0 0 1 1 1/2 −1/2 1 −1 0 1/2 1/2 1 1 0 −1/2 1/2 1/2 − 1/2 −1/2 + 1/2 1/2 + 1/2 1/2 − 1/2 1/2 −1/2 1/2 + 1/2 0 0 1 0 0 1 5.2 Matrix Equations 297 Elementary Matrices. An elementary matrix E is the result of applying a combination, multiply or swap rule to the identity matrix. It follows from the definition of matrix multiplication that EA is the result of the corresponding combination, multiply or swap rule applied to matrix A. The computer algebra system maple easily writes out typical 4 × 4 elementary matrices (C=Combination, M=Multiply, S=Swap) as follows. with(linalg): Id:=diag(1,1,1,1); C:=addrow(Id,2,3,c); M:=mulrow(Id,3,m); S:=swaprow(Id,1,4); with(LinearAlgebra): Id:=IdentityMatrix(4); C:=RowOperation(Id,[3,2],c); M:=RowOperation(Id,3,m); S:=RowOperation(Id,[4,1]); The reader is encouraged to write out the above elementary matrices by hand or machine, obtaining C= 1 0 0 0 0 1 c 0 0 0 1 0 0 0 0 1 , M = 1 0 0 0 0 0 0 1 0 0 0 m 0 0 0 1 , S = 0 0 0 1 0 1 0 0 0 0 1 0 1 0 0 0 . Constructing Elementary Matrices E. Multiply Change a one in the identity matrix to symbol m 6= 0. Combination Change a zero in the identity matrix to symbol c. Swap Interchange two rows of the identity matrix. Constructing E −1 from elementary matrix E. Multiply Change diagonal multiplier m 6= 0 in E to 1/m. Combination Change multiplier c in E to −c. Swap The inverse of E is E itself. Theorem 8 (The rref and elementary matrices) Let A be a given matrix of row dimension n. Then there exist n × n elementary matrices E1 , E2 , . . . , Ek such that rref (A) = Ek · · · E2 E1 A. The result is the observation that left multiplication of matrix A by elementary matrix E gives the answer EA for the corresponding multiply, combination or swap operation. The matrices E1 , E2 , . . . represent the multiply, combination and swap operations performed in the frame sequence which take the First Frame into the Last Frame, or equivalently, original matrix A into rref (A). 298 Illustration. Consider the following 6-frame sequence. 1 2 3 A1 = 2 4 0 3 6 3 Frame 1, original matrix. 1 2 3 A2 = 0 0 −6 3 6 3 Frame 2, combo(1,2,-2). 1 2 3 A3 = 0 0 1 3 6 3 Frame 3, mult(2,-1/6). 1 2 3 1 A4 = 0 0 0 0 −6 Frame 4, combo(1,3,-3). 1 2 3 A5 = 0 0 1 0 0 0 Frame 5, combo(2,3,-6). 1 2 0 A6 = 0 0 1 0 0 0 Frame 6, combo(2,1,-3). Found rref . The corresponding 3 × 3 elementary matrices are 1 0 0 E1 = −2 1 0 0 0 1 Frame 2, combo(1,2,-2) applied to I. 1 0 0 E2 = 0 −1/6 0 0 0 1 1 0 0 E3 = 0 1 0 −3 0 1 Frame 4, combo(1,3,-3) applied to I. 1 0 0 1 0 E4 = 0 0 −6 1 Frame 3, mult(2,-1/6) applied to I. Frame 5, combo(2,3,-6) applied to I. 1 −3 0 1 0 E5 = 0 0 0 1 Frame 6, combo(2,1,-3) applied to I. Because each frame of the sequence has the succinct form EA, where E is an elementary matrix and A is the previous frame, the complete frame sequence can be written as follows. 5.2 Matrix Equations 299 A2 = E1 A1 Frame 2, E1 equals combo(1,2,-2) on I. A3 = E2 A2 Frame 3, E2 equals mult(2,-1/6) on I. A4 = E3 A3 Frame 4, E3 equals combo(1,3,-3) on I. A5 = E4 A4 Frame 5, E4 equals combo(2,3,-6) on I. A6 = E5 A5 Frame 6, E5 equals combo(2,1,-3) on I. A6 = E5 E4 E3 E2 E1 A1 Summary, frames 1-6. This relation is rref (A1 ) = E5 E4 E3 E2 E1 A1 , which is the result of the Theorem. The summary is the equation 100 1 00 100 1 −3 0 1 0 0 rref (A1 ) = 0 1 00 1 0 0 1 00 − 61 0−2 1 0 A1 001 0 0 1 0 −6 1 −3 0 1 0 0 1 The inverse relationship A1 = E1−1 E2−1 E3−1 E4−1 E5−1 rref (A1 ) is formed by the rules for constructing E −1 from elementary matrix E, page 297, the result being 100 1 00 100 100 130 A1 = 2 1 00 −6 00 1 00 1 00 1 0 rref (A1 ) 001 0 01 301 061 001 1 Example (Reduced Row–Echelon Form) Find the reduced row–echelon form of the coefficient matrix A using the rref method, page 292. Then solve the system. x1 + 2x2 − x3 + x4 = 0, x1 + 3x2 − x3 + 2x4 = 0, x2 + x4 = 0. Solution: The coefficient matrix A and its rref are given by (details below) 1 A= 1 0 −1 1 −1 2 , 0 1 2 3 1 1 rref (A) = 0 0 0 1 0 −1 −1 0 1 . 0 0 Using variable list x1 , x2 , x2 , x4 , the equivalent reduced echelon system is − x3 x1 x2 − x4 + x4 = 0, = 0, which has lead variables x1 , x2 and free variables x3 , x4 . The last frame algorithm applies to write the standard general solution. This algorithm assigns invented symbols t1 , t2 to the free variables, then backsubstitution is applied to the lead variables. Then x1 x2 x3 x4 = = = = t1 + t2 , −t2 , t1 , t2 , −∞ < t1 , t2 < ∞. 300 Details ∗ 1 1 0 1 0 0 1 0 0 of the rref method. 2 −1 1 3 −1 2 1 0 1 2 −1 1 1∗ 0 1 1 0 1 0 −1 −1 1 0 1 0 0 0 The coefficient matrix A. Leading one identified, marked as 1∗ . Apply the combination rule to zero the other entries in column 1. Mark the row processed. Identify the next leading one. Apply the combination rule to zero the other entries in column 2. Mark the row processed; rref found. Proof of Theorem 6: Assume AB = I. Let C = BA − I. We intend to show C = 0, then BA = I, which completes the proof. Compute AC = ABA − A = AI − A = 0. It follows that the columns ~y of C are solutions of the homogeneous equation A~y = ~0. The proof is completed by showing that the only solution of A~y = ~0 is ~y = ~0, because then C has all zero columns, or equivalently, C = 0. First, B~u = ~0 implies ~u = AB~u = A~0 = ~0, hence B has an inverse, and then B~x = ~y has a unique solution ~x = B −1 ~y . Suppose A~y = ~0. Write ~y = B~x. Then ~x = I~x = AB~x = A~y = ~0. This implies ~y = B~x = B~0 = ~0. The proof is complete. Proof of Theorem 7: Details for (a). Let C = aug(A, I) and assume rref (C) = aug(I, B). Solving ~ = ~0 is equivalent to solving the system AY ~ + IZ ~ = ~0 the n × 2n system C X ~ and Z. ~ This system has exactly the same solutions with n-vector unknowns Y ~ + BZ ~ = ~0, by the equation rref (C) = aug(I, B). The latter is a as I Y ~ and reduced echelon system with lead variables equal to the components of Y ~ Multiplying by A gives AY ~ + free variables equal to the components of Z. ~ = ~0, hence −Z ~ + AB Z ~ = ~0, or equivalently AB Z ~ =Z ~ for every vector Z ~ AB Z ~ be a column of I shows (because its components are free variables). Letting Z that AB = I. Then AB = BA = I by Theorem 6, and B is the inverse of A. Details for (b). Assume AB = I. We prove the identity rref (aug(A, I)) = ~ + IZ ~ = ~0 have a solution Y ~ , Z. ~ Multiply by B aug(I, B). Let the system AY ~ ~ ~ ~ ~ ~ to obtain BAY + B Z = 0. Use BA = I to give Y + B Z = 0. The latter system ~,Z ~ as a solution. Conversely, a solution Y ~,Z ~ of Y ~ +B Z ~ = ~0 is a therefore has Y ~ ~ ~ solution of the system AY + I Z = 0, because of multiplication by A. Therefore, ~ + IZ ~ = ~0 and Y ~ + BZ ~ = ~0 are equivalent systems. The latter is in reduced AY row-echelon form, and therefore rref (aug(A, I)) = aug(I, B). Details for (c). Suppose not, then A is invertible. Then (b) implies aug(C, B) = rref (aug(A, I)) = aug(I, B), which gives C = I, a contradiction. 5.2 Matrix Equations 301 Exercises 5.2 0 matrix A, assume a homogeneous sys- 11. 0 0 ~ = ~0 with variable list x1 , . . . , tem AX 0 xn . List the lead and free variables. Then report the rank and nullity of 1 matrix A. 0 12. 0 0 1 3 0 0 0 1. 0 0 0 1 0 0 0 0 0 0 Lead and free variables. For each 0 1 0 0 0 2. 0 0 1 0 3 0 0 0 1 2 1 0 0 0 0 1 0 0 5 2 0 0 0 1 0 0 3 0 0 0 0 1 0 0 0 0 1 0 0 0 1 0 Elementary Matrices. Write the 3×3 elementary matrix E and its inverse E −1 for each of the following operations (defined in the text). 13. combo(1,3,-1) 0 1 3 0 3. 0 0 0 1 0 0 0 0 14. combo(2,3,-5) 15. combo(3,2,4) 1 2 3 0 4. 0 0 0 1 0 0 0 0 1 0 5. 0 0 2 0 0 0 16. combo(2,1,4) 17. combo(1,2,-1) 18. combo(1,2,-e2 ) 3 0 0 0 19. mult(1,5) 20. mult(1,-3) 21. mult(2,5) 1 1 0 6. 0 0 1 0 0 0 22. mult(2,-2) 1 1 3 5 0 7. 0 0 0 0 1 0 0 0 0 0 1 2 0 3 4 8. 0 0 1 1 1 0 0 0 0 0 23. mult(3,4) 24. mult(3,5) 25. mult(2,-π) 26. mult(2,π) 27. mult(1,e2 ) 0 0 9. 0 0 0 0 0 0 1 0 0 0 2 0 0 0 0 1 0 0 28. mult(1,-e−2 ) 0 0 0 0 0 0 0 0 1 0 0 0 1 0 0 0 31. swap(2,3) 0 0 10. 0 0 29. swap(1,3) 30. swap(1,2) 32. swap(2,1) 33. swap(3,2) 302 34. swap(3,1) Elementary Matrix Properties. For each given matrix A, and operation Frame Sequences and Elementary (combo, swap, mult), write the reMatrices. Given the final frame B of sult B of performing the operation on a sequence starting with matrix A, and the given operations, find matrix A. 1 1 0 35. B = 0 1 2 , operations 0 0 0 combo(1,2,-1), combo(2,3,-3), mult(1,-2), swap(2,3). 1 1 0 36. B = 0 1 2 , operations 0 0 0 combo(1,2,-1), combo(2,3,3), mult(1,2), swap(3,2). 1 1 2 37. B = 0 1 3 , operations 0 0 0 combo(1,2,-1), combo(2,3,3), mult(1,4), swap(1,3). 1 1 2 38. B = 0 1 3 , operations 0 0 0 combo(1,2,-1), combo(2,3,4), mult(1,3), swap(3,2). A. Then compute the elementary matrix E for the operation and verify B = EA. 1 1 39. , mult(2,1/3). 0 3 1 1 0 2 3 , mult(1,3). 0 1 1 0 2 1 , combo(3,2,-1). 1 1 40. 0 0 1 41. 0 0 1 0 3 1 1 43. 0 0 1 1 0 42. 44. 1 0 3 1 , combo(2,1,-3). 2 3 , swap(2,3). 1 , swap(1,2).